Vol. 82 (1992) ACTA PHYSICA POLONICA Α Νo 2

Proceedings of the ISSSRNS ,92, Jaszowiec 1992

EXTENDED X-RAY ABSORPTION FINESTRUCTURE IN CATALYSIS

A.J. RENOUPREZ

Institut de Recherches sur la Catalyse2, av. Albert Einstein, 69626 Villeurbanne cedex, France

Applications of extended X-ray absorption fine structure and X-ray edgestructure to catalysis are described. Special attention is paid to the studyof supported mono- and bimetallic catalysts. The method used to collect insitu data, under reaction conditions are also considered.

PACS numbers: 61.10.Lx, 82.65.Jv

1. Introduction

During the last ten years extended X-ray absorption fine structure (EXAFS)and related X-ray absorption methods have proved to be powerful tools to charac-terize catalytic materials. Two reasons can be invoked. First, it is simultaneouslya spectroscopic and a stuctural probe. Compared to diffraction, it has the uniqueadvantage that one can know around which atom the local order is determined.Second, it has a high sensitivity, largely ascribable to the considerable photon fluxof the modern synchrotrons.

Among the properties of EXAFS which makes it useful for the study ofcatalysts are:

— long-range order is not required and most of the catalytic systems do notcontain well-ordered phases,

— the local atomic arrangement can be determined around each type of atomseparately by tuning the photon energy,

— the structural information is obtained by a simple analysis of the data withan accuracy comparable to that obtained by diffraction on single crystals,

— the data collection is so fast that kinetic measurements are now currentlyperformed,

— in situ measurements under reaction conditions can be performed.

(295)

296 A.J. Renouprez

2. Theory of EXAFS and data analysis

2.1. Theory

The fundamental dscovery of EXAFS was actually performed in 1971 byStern and Sayers [1] who showed that the oscillations of the X-ray absorptioncoefficient located at an energy higher than the absorption edge of a given elementare related to the short-range order around this element.

They derived a formula which expresses that the Fourier transform of theEXAFS signal is a radial distribution function which peaks at distances corre-sponding to the successive coordination shells of atoms.

When X-ray photons have enough energy to free a bound electron in theatom, an absorption edge is observed: K- or L-edges for 1s or 2p electrons. Thekinetic energy of the ejected photoelectron is the difference between the photonenergy and the binding energy of electron.

For photon energies 15-20 eV above the edge, only the final state of thephotoelectron is perturbed by the surroundings. This final state is modified by asingle scattering from each surrounding atom and is the superposition of outgoingand scattered waves. This phenomenon is represented in Fig. 1.

The outgoing wave has an amplitude proportional to exp(ikr)/r, wherek = 2π/λ and λ is the wavelength. The amplitude of the backscattered wave,characteristic of the backscattering atom, reads at ri, position of this atom,

At the origin (r = 0) the wave has travelled the distance 2ri and its amplitude isproportional to

Actually, the electrons do not travel in a constant potential and both centraland scattering atoms add an additional phase shift to the wave and the amplitudebecomes

Extended X-ray Absorption Fine Structure in Catalysis 297

The interference between backscattered and outgoing waves modifies theabsorption coefficient χi(k):

the real part of (3).Two other factors modify the absorption coefficient. First, the photoelectron

has a finite lifetime because of scattering from surrounding atoms. It is taken intoaccount by a mean free path term exp(-2ri/L), where L is a mean free path ofthe photoelectron, in general a function of k. Also in a coordination shell, all theatoms will not be at the same distance from the central atom, because of thermalvibrations and static disorder. If this disorder has a Gaussian distribution, it willadd a factor exp(-2k 2σ2i ), where σ is the relative mean square deviation from thedistance ri. The final EXAFS formula becomes, when extended to all the atomsof the various shells,

A Fourier transform of X(k) will thus lead to the number of atoms Νi inthe shell located at Ri from the central atom. The nature of these atoms can beidentified because A(k) is characteristic of a given backscattering element.

2.2. Data analysis

The experimental absorption coefficient μ(Ε) has to be normalized first tothe atomic absorption μ0(Ε) to obtain X(Ε):

This subtraction is performed by approximating first the experimental databy a simple polynomial followed by digital high-pass filtering. The resulting curveis then subtracted from the experimental one. The successive steps of the datareduction are shown in Fig. 2.

The second step consists in converting X(Ε) to X(k) with

where Ε0 is the energy of the edge. Note that the choice for Ε0 will have a markeinfluence on Ri , the radius of a shell.

The phase shifts contained in expression (5) were computed for a given pairof atoms by McKale [2] together with the backscattering amplitude of all thyelements. One can also make use of experimental parameters measured on a compound of known stucture.

The fit of expression (5) to the experimental data should thus lead to 11and R for the various shells; however, N and σ (the Debye—Waller term) are noindependent variables. Since the thermal part of σ decreases at low temperature

298 Α.J. Renouprez

a more accurate determination of N would be obtained by low-temperature ex-periments.

Usually one isolates first each shell on the Fourier transform of the EXAFSsignal and adjusts it to a model. If the statistics is of the order of 10 -4 and if thedata extends over a large k range (0-18 Å -1 ), the accuracy of R is ±0.02 Å and15% of N.

An excellent review on the various problems of data analysis is presented inRef. [4].

3. Experimental

3.1. Synchrotron radiation

At the synchrotron facility, the photons are produced by high-energy elec-trons or positrons (500 MeV to 6 GeV) travelling in a storage ring.

Extended X-ray Absorption Fine Structure in Catalysis 299

The broad wavelength distribution has a maximum of intensity close to acritical value

where Rm is the radius of curvature of the bending magnet and E — the energyof the electrons in GeV. The total scattered power Ρ is

The interesting features of the synchrotron radiation are:— a flux 5-6 orders of magnitude larger than with X-ray tubes, typically 10 14

photons.s-1•mrad-2 for 0.1% band width,— a large range of energies: 2-50 keV at the European Facility (ESRF),

— a high brilliance resulting from the small source size (0.1-0.2 mm) and lowvertical divergence of the beam (0.1 mrad) . .

This last point, together with the width of the rocking curve of the monochromator,determines the energy resolution.

3.2. The instrument

The EXAFS instument is composed of three main parts:1. The X-ray. optics.2. The detection system.3. The sample environment.

The X-ray optics is composed of a mirror which delivers a parallel beamand removes the high-energy photons when necessary. The mirror is followed by atwo crystal monochromator which is rotated to scan the energies. Silicon crystalswith (111), (200) or (311) orientations are generally used; they have an energyresolution of 2 x 10 -4 to 10 -5 .

The detection system is composed of a first counter which measures I 0 , thephoton flux before the sample, and a second detector measuring I, the intensitybehind it. The absorption coefficient is

Ι and ι0 are usually measured by ion chambers, a type of counter limited at lowflux by its noise and at high flux by charge accumulation effects resulting in a lossof linearity.

As a function of the available flux, one has to vary the nature of the gas inthe counters and its pressure to optimize the efficiency.

Teo [3] showed that the signal to noise ratio (S/N) is given for thick samplesby the expression

300 A.J. Renouprez

where μT = μA + μB, μA i8 the absorption coefficient of the studied element, μB— of the matrix, ΔμΑ/μΑ is the EXAFS signal. The optimum sample thickness,Χ, occurs at μTΧ = 2.56. For thin samples

It can easily be seen that absorption measurements can be difficult if μΒΧBis large, i.e. if for example the catalyst is supported on a highly absorbing material.This is the case if the active element has a low Z number and thus an edge locatedat low energy. It will be for example quite impossible to study by transmissiondiluted samples composed of Fe or Cr supported on ZrO2. On the contrary, it willbe easy with Pd/SiO2 , since the Pd K-edge is located at 24 keV, an energy whereμB for SiO2 is low.

Typically, with a photon flux of 10 12 photons/s, ΔμΑ/μΑ 10-2 , μΑ/μT =10 -3 , the signal to noise ratio will be of the order of 10 2 in 1 s. Thus, the time ofexperiment for a 1 keV scan and 2 eV step is of the order of 10 min.

Many effects worsen the results and, first, the nonlinearity and saturation ofthe ion chambers. For this reason photodiodes are now preferred; they have a lowdark current and remain linear for photon flux larger by two orders of magnitudethan the gas detectors.

For the study of diluted elements a good alternative to transmission detec-tion is fluorescence. Indeed, the excited absorber atom with a core hole, relaxesby undergoing a radiative transition. The fluorescence yield near the edge is pro-portional to the absorption coefficient. This fluorescence yield varies with Z, forexample from 0.35 for Fe to 0.96 for Pt.

For example, an EXAFS experiment at the Fe K-edge located at 7.112 keV isperformed by collecting the photons corresponding to K α1 , K, 2, and Kβ emissionslocated at 6.403, 6.39 and 7.057 keV. To improve the S/N in fluorescence, one hasto remove the elastic photons. A good solution is to use a filter composed of theZ-1 element. In the above example, the elastic photons have an energy comprised,during the scan, between 7.1 and 8 keV and a Mn filter with a Kedge locatedat 6.54 keV will be used. Scintillation detection equipped with plastic scintillatorscan collect up to 3 x 10 6 photons/s. One also uses solid state detection which havea good energy resolution (200 eV at 10 keV) but are limited in counting rate to104 imp./s. The multicell Ge detector manufactured by Camberra is composed of13 individual diodes. It has an energy resolution of 160 eV at 6 keV and a countingrate close to 10 5 imp./s.

Less expensive solutions will certainly appear soon. They will be composed ofmultistrip Si or Ge detectors with 100 to 200 strips with fully integrated electronics.

3.3. Sample environment

Catalytic systems are generally composed of finely divided metals or sul-phides which can undergo transformations when exposed to atmosphere. More-over, it is highly desirable to investigate stucture modifications, poisoning etc.under catalytic conditions. For these reasons, many authors have designed high--temperature cells [4] where the catalysts can be heated up to 900 K under gas

Extended X-ray Absorption F ine Structure in Catalysis 301

flow and cooled down to 100 K to minimize the Debye-Waller factor. Some ofthem can work both in transmission and fluorescence mode.

In a recent achievement [8], such a cell was coupled with a chromatographicdetection to study under pressure (10 bar) the cocking of a catalyst during areforming reaction.

4. EXAFS on metallic catalysts

4.1. Monometallic catalysis

Α large number of studies have been performed on metallic supported cat-alysts during the last ten years; this interest is due to the importance of the oilreforming catalytic process. EXAFS was able to elucidate the various steps of thecatalysts preparation and of their deactivation under catalytic conditions.

Α good example is the study of the preparation of Pd/alumina catalystsknown to be selective in the hydrogenation of dienes to alcenes.

Lesage [5] showed that when α-alumina is impregnated with (C5H7O2)2Pd,the Pd atom is surrounded by 4 oxygen atoms located at 1.96 Α. After drying at100°C, one finds 5 oxygen atoms at 1.98 Å; this corresponds to the loss of oneACAC molecule and to the coordination of the Pd-C5Η7O2 residue to 3 oxygenatoms of the alumina. Calcination up to 700°C progressively leads to the formationof oxide and the observation of Pd-Pd pairs in the second coordination shell.

Α large number of studies were also devoted to the study of reforming cata-lysts composed of Pt/γ-alumina [6, 7]. Special attention was paid to the influenceof chlorine on the formation of the catalysts.

In the study by Guyot-Sionnest [8], a non-chlorinated and a chlorinatedcompound (1% Pt, 2% Cl) were compared. It was shown that after the calcina-tion treatment in air at 500°C for non-chlorinated samples, the first coordinationsphere is that of β-PtO2 with 6 oxygen atoms surrounding platinum. But in thesecond and third shells composed of Pt—Pt pairs at 3.2 and 3.5 Å, the coordinationnumbers are 1/2 those of PtO2. This indicates that the oxide particles have a verysmall size. When Cl is present in the sample, Cl atoms are observed in the firstshell and the numbers of Pt atoms in the second and third shell are still reduced.

During catalyst reduction in flowing H2 , the number of Pt-O begins to de-crease at 150°C to reach a stable value of 0.3 at 460°C. Simultaneously, Pt-Pt firstneighbours appear above 200°C to reach 5.4 for Pt/Al2O3 and 4.1 for Pt/Al2O3—Clat 460°C.

Such low coordination numbers of 4 to 5.5 compared to 12 in bulk materialsmean that the particles are composed of 10-15 atoms. The comparison of theFourier transform (FT) of PtO2 and of the catalysts is shown in Fig. 3.

An important point should be outlined: in this type of study, precise valuesfor the backscattering amplitude and phase shift parameters are needed. For this,model compounds with known crystallographic parameters are studied by EXAFS.For the Pt-O pair β-PtO2 was used, for the Pt-Cl, Η2PtC16 and a Pt foil for thePt-Pt pair.

302 A.J. Renouprez

Actually, these in situ experiments performed at high temperature sufferfrom a limited k range, typically 10 A -1 ; the consequence is that the accuracy ofR is certainly not larger than 0.05 Å.

A catalytic conversion of N-heptane was also performed on these catalystsin an EXAFS cell. During the various reactions occurring at 350°C, the formationof Pt—C bonds are observed. After the reaction, Pt is surrounded by 1.4 C atomsat 1.96 A and 5 Pt atoms at 2.67 Å. Flowing hydrogen at 4500C for 3 h removesa large fraction of the carbon (0.4 Pt-C only). The successive transformations ofthe catalyst are illustrated in Fig. 4.

Another example of a complete stucture determination of a metallic sup-ported catalyst was performed for Pd/Y-zeolite [6]. These 10 Å particles encagedin the large cavity of the zeolite were studied at 100 K. Because of the low temper-ature, the data extend up to 18 Å-1 and the first three layers could be analysed(Figs. 5 and 6, Table I). It was concluded that the particles are composed of 45±5atoms arranged in tuncated octahedron with fcc symmetry.

Extended X-ray Absorption Fine Structure in Catalysis 303

304 A.J. Renouprez

Extended X-ray Absorption Fine Structure in Catalysis 305

4.2. Bimetallic catalysts

Modern reforming catalysts are now composed of platinum in associationwith other elements such as Re, Sn, Ge, Ir, etc. The characterization of smallbimetallic particles requires a probe with an atomic resolution. One needs to an-swer the following questions:

— Are the two metals fully reduced?— Is a tue alloy formed by these two metals?— Is there any surface segregation?

Obviously, the complete answer necessitates experiments at both edges. Theanalysis of the EXAFS data is more complex than in the previous case because,generally, model compounds containing both elements are not available.

In the well-known case of Ru-Cu/SiO2 studied by Simfeld [9] (Fig. 7) thephase shifts δCu-Ru and δRu-Cu are obtained by measuring, first, Ru and Cu edgesof RuO2 and CuO and then,

and

It was found that the coordination numbers in the first shell are, respectively,9 and 11 for Cu and Ru. This indicates that Cu with lower coordination numbermigrates at the surface of the particle.

However, the chemical analysis corresponds to a 1 :1 ratio of Cu to Ru butone finds 92% of Ru atoms around Ru and 50% Cu atoms around Cu with distancesRu-Ru = 2.65 Α. on Cu-Cu = 2.58 Α. This confirms that Ru atoms are in thecore of the particle with mainly Ru atoms around them and that Cu has a mixedneighbourhood with Ru atoms probably lying in the second layer below the surface.

These observations are in line with oxygen chemisorption measurementswhich show that only Cu is oxidized and that the CuO layer protects the un-derlying Ru atoms.

306 A.J. Renouprez

Extended X-ray Absorption Fine Structure in Catalysis 307

White lines observed at adsorption edges are attributed to transition from2p to nd states. For elements with uncompleted bands, the photoelectron is ejectedto empty levels close to the Fermi level, resulting in an increase in the adsorptioncoefficient, a few eV above the edge. This was used by many authors [10] to probethe chemical state of this element. Upon exposure to oxygen, Pt, Ir, Rh exhibita strong increase in the LIII line compared to the LII which can be related tocreation of vacancies.

This type of probe can also be used to investigate more subtle electronicstructure modifications which occur in bimetallic systems. Α good example is thestudy of Pt Lm-edge of Pt-Fe bimetallic catalysts performed by Moraweck etal. [11].

It was shown that upon increasing the Fe concentration the 5d holes in Ptbands are progressively filled by iron electrons. This can be related to modificationsof the catalytic properties. .

5. Hydrodesulphurization catalysts

The study of hydrodesulphurization catalysts used in the treatment of cudeoils constitutes the best example of the success of X-ray adsorption in the structuredetermination of complex systems [12, 13]. They are constituted of molybdenumoxides supported on alumina associated with a promoter composed of cobalt ox-ides.

In the oxidized state, the Co-Mo/Al2O3 catalyst contains cobalt atoms sur-rounded by 4 oxygen atoms at 1.99 Α like in CoAl2O4 and Mo is surrounded by 6oxygen atoms at 1.81 Α.

Upon sulphurization of the catalyst, Co is now surrounded by 4 O at 2.03 Αand 6 S at 2.33 Α. The Co—O distance is too short compared to that found inCoO (2.14 Å) and too large for a Co in tetrahedral sites as in CoAl2O4 (Table II).Actually, this can be interpreted by the presence of two types of cobalt species;the first one has a coordination of 6 as in CoS, the second one with 4 O atomswould anchor the particle to Al2O3.

The investigations at the Mo edge showed that Mo is surrounded by 6 S at2.37 Ǻ as in MoS2 but in the second sphere only 4 Mo atoms at 3.25 Α are found

308 A.J. Renouprez

instead of 6in MoS2. This is the indication of the presence of small (10 Å) MoS2clusters.

The promoting effect of Co in these catalysts can be attributed to an electrontransfer from Co-S bonding orbitals to antibonding Mo-S. An increase in electrondensity on Mo-S would enhance the activation of H2.

6. Conclusions

X-ray absorption is now one of the most powerful analytical tool to studythe stucture of catalysts. Moreover, the last versions of quick EXAFS or disper-sive EXAFS are able to follow the kinetics of structural modifications occurringduring catalytic reactions. Its ability to probe the various elements constitutingthe catalyst is also a precious help to optimize its preparation.

References

[1] D.E. Sayers, Ε.Α. Stern, F.W. Lytle, Phys. Rev. Lett. 27, 1204 (1971).

[2] A.G. McKale, B.W. Veal, A.P. Paulikas, S.K. Chan, G.S. Knapp, J. Am. Chem.Soc. 110, 3763 (1988).

[3] B.K. Teo, in: EXAFS Spectroscopy, Eds. B.K. Teo, D.C. Joy, Plenum Press, NewYork 1979, p. 13.

[4] X-Ray Adsorption, Chemical Analysis, Vol. 92, Eds. D.C. Koningsberger, R. Prins,Wiley, New York 1988.

[5] E. Lesage-Rosenberg, Ph.D. thesis, Technip, 1984.[6] B. Moraweck, A. Renouprez, Surf. Sci. 106, 35 (1981).[7] B. Moraweck, G. Clugnet, A. Renouprez, J. Chim. Phys. 83, 265 (1986).[8] N. Guyot-Sionnest, Ph.D. thesis, LURE, Orsay 1991.[9] T.H. Sinfeld, G.H. Via, R.B. Greegor, J. Chem. Phys. 75, 5527 (1981).

[10] F.W. Lytle, J. Catal. 43, 376 (1976).[11] B. Moraweck, P. Bondot, D. Goupil, P. Fouilloux, A. Renouprez, J. Phys. (France)

8, 39, 263 (1986).[12] G. Sankar, S. Vasudevan, C.N.R. Rao, J. Phys. Chem. 91, 2011 (1987).[13] B.S. Clausen, H. Topsoe, R. Candia, T. Villadsen, B. Lengeler, T. Alsnielsen,

Τ. Christensen, J. Phys. Chem. 85, 3868 (1981).